Presentation
in
Competition: Auto-Bidding in Large-Scale Auctions: Learning Decision-Making in Uncertain and Competitive Games
General Track Winner (MingGeZheJiu) Presentation
Qingmao Yao · Haoran Sun · Tingting Meng · Hao Xu · Chu Xu
Abstract:
In this auto-bidding competition, we develop a time-varying, risk-avoiding linear bidding strategy that achieves a score of 0.5635 in the final rounds. Building on prior work, we first establish a general bidding model for Target CPA bidding. By leveraging the cyclic and temporal patterns of pValue, we introduce time-varying parameterization. Finally, we optimize the bidding parameters in a simulated offline environment and incorporate online feedback for fine-tuning, ultimately reaching a score of 0.5635.
Chat is not available.